har0ke/FFmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

avcodec/fft: Add revtab32 for FFTs with more than 65536 samples

Browse Source 
x86 optimizations are used only for the cases they support (<=65536 samples)

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
This commit is contained in:
Michael Niedermayer 2016-03-04 15:39:55 +01:00
parent ae76b84221
commit 305344d89e
3 changed files with 30 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
libavcodec/fft.h
View File

@ -110,6 +110,7 @@ struct FFTContext {
void (*mdct_calcw)(struct FFTContext *s, FFTDouble *output, const FFTSample *input);
enum fft_permutation_type fft_permutation;
enum mdct_permutation_type mdct_permutation;
uint32_t *revtab32;
};
#if CONFIG_HARDCODED_TABLES

31
libavcodec/fft_template.c
View File

@ -143,14 +143,23 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
{
int i, j, n;
s->revtab = NULL;
s->revtab32 = NULL;
if (nbits < 2 || nbits > 17)
goto fail;
s->nbits = nbits;
n = 1 << nbits;
s->revtab = av_malloc(n * sizeof(uint16_t));
if (!s->revtab)
goto fail;
if (nbits <= 16) {
s->revtab = av_malloc(n * sizeof(uint16_t));
if (!s->revtab)
goto fail;
} else {
s->revtab32 = av_malloc(n * sizeof(uint32_t));
if (!s->revtab32)
goto fail;
}
s->tmp_buf = av_malloc(n * sizeof(FFTComplex));
if (!s->tmp_buf)
goto fail;
@ -192,16 +201,22 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
fft_perm_avx(s);
} else {
for(i=0; i<n; i++) {
int k;
j = i;
if (s->fft_permutation == FF_FFT_PERM_SWAP_LSBS)
j = (j&~3) | ((j>>1)&1) | ((j<<1)&2);
s->revtab[-split_radix_permutation(i, n, s->inverse) & (n-1)] = j;
k = -split_radix_permutation(i, n, s->inverse) & (n-1);
if (s->revtab)
s->revtab[k] = j;
if (s->revtab32)
s->revtab32[k] = j;
}
}
return 0;
fail:
av_freep(&s->revtab);
av_freep(&s->revtab32);
av_freep(&s->tmp_buf);
return -1;
}
@ -210,15 +225,21 @@ static void fft_permute_c(FFTContext *s, FFTComplex *z)
{
int j, np;
const uint16_t *revtab = s->revtab;
const uint32_t *revtab32 = s->revtab32;
np = 1 << s->nbits;
/* TODO: handle split-radix permute in a more optimal way, probably in-place */
for(j=0;j<np;j++) s->tmp_buf[revtab[j]] = z[j];
if (revtab) {
for(j=0;j<np;j++) s->tmp_buf[revtab[j]] = z[j];
} else
for(j=0;j<np;j++) s->tmp_buf[revtab32[j]] = z[j];
memcpy(z, s->tmp_buf, np * sizeof(FFTComplex));
}
av_cold void ff_fft_end(FFTContext *s)
{
av_freep(&s->revtab);
av_freep(&s->revtab32);
av_freep(&s->tmp_buf);
}

3
libavcodec/x86/fft_init.c
View File

@ -26,6 +26,9 @@ av_cold void ff_fft_init_x86(FFTContext *s)
{
int cpu_flags = av_get_cpu_flags();
if (s->nbits > 16)
return;
#if ARCH_X86_32
if (EXTERNAL_AMD3DNOW(cpu_flags)) {
/* 3DNow! for K6-2/3 */